What are the Microsoft responsible AI principles and why they are relevant to you as an AI practitioner?
- [Instructor] Let's now talk about Microsoft Responsible AI principles. If you are a pop culture fan like me you can probably name at least three major movie franchises. The talk about our artificial intelligence going rogue and almost destroying humanity, right? Well, true it's unlikely that robots will come back in time anytime soon to kill the last savior of the human race. However, the fact is that there are plenty of recent examples of both companies and governments taking morally dubious decisions when it comes to AI in people's data. Microsoft is aware of this potential for misuse and once you set higher standards when it comes to AI solutions. The Microsoft Responsible AI code is composed of six main principles. Fairness, AI Systems should treat people fairly by enforcing mechanisms to prevent biases that could disadvantage a certain group be that based on race, religion or just ability. This is especially important in areas such as criminal justice or banking and finance but any industry can be in fact impacted by bias. Reliability and safety, it's important to make sure that AI Systems operate reliably, safely and consistently independently of anticipated conditions. This has high relevance scenarios such as medicine or self driving vehicles but it's important to thoroughly test the AI Systems independently of whether they could pose a physical threat to human beings. Privacy and security, With the advent of privacy standards such as GDPR it's essential that AI systems respect individual privacy and protect customers' data at all costs. This has been probably the most violated principle in recent years, be that by weak corporate security practices or by intention in misuse of customer data by governments or private companies. Inclusiveness, our planet has over 1 billion people with some sort of disability. An AI can offer an immense help to them, therefore, it's important to design AI solutions with inclusive design practices so that everyone can benefit from it. Transparency, the very nature of AI as you see in later modules might give rise to a black box effect where you don't exactly know why the AI is giving those results. It's important therefore to understand the reasoning behind a certain AI model to be able to proactively fight bias or exclusion practices. Accountability, this principle defense that humans are ultimately responsible for the decisions of AI Systems. Even highly automated ones, it must be able to override their decisions if needed. These principles are quite important to Microsoft and therefore to you as a test taker you should expect at least a couple of questions related to this principles on the exam.